-
Notifications
You must be signed in to change notification settings - Fork 0
/
loss.py
executable file
·62 lines (46 loc) · 2.11 KB
/
loss.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
import torch
import torch.nn as nn
import torch.nn.functional as F
from functools import partial
import numpy as np
def min_mask_loss(mask, min_mask_coverage):
return F.relu(min_mask_coverage - mask.mean(dim=(1, 2, 3))).mean()
def min_permask_loss(mask, min_mask_coverage):
return F.relu(min_mask_coverage - mask.mean(dim=(2, 3))).mean()
def min_mask_loss_batch(mask, min_mask_coverage):
return F.relu(min_mask_coverage - mask.mean())
def binarization_loss(mask):
return torch.min(1-mask, mask).mean()
def create_circle_mask(h, w, center=None, radius=None):
if center is None:
center = (int(w/2), int(h/2))
if radius is None:
radius = min(center[0], center[1], w-center[0], h-center[1])
Y, X = np.ogrid[:h, :w]
dist_from_center = np.sqrt((X - center[0])**2 + (Y-center[1])**2)
mask = dist_from_center <= radius
return mask.astype(int)
class MaskLoss:
def __init__(self, min_coverage=0.35, coef_binary=5, coef_area=5, min_coef_binary=0.5, decay_length=5000, device='cpu'):
self.coef_binary = coef_binary
self.coef_area = coef_area
self.phi1 = min_coverage
self.min_coef_binary = min_coef_binary
self.decay_length = decay_length
def __call__(self, M, current_iter, mask=None):
if mask == None:
loss_binary = torch.min(1-M, M).mean()
else:
loss_binary = (torch.min(1-M, M)*mask).mean()
loss_area = F.relu(self.phi1 - M.mean(dim=(-2,-1))).mean()
return max((self.min_coef_binary - self.coef_binary) / self.decay_length * current_iter + self.coef_binary, self.min_coef_binary) * loss_binary \
+ self.coef_area * loss_area
class FineMaskLoss:
def __init__(self, min_coverage=0.0, max_coverage=0.01, coef_m_fine=5, device='cpu'):
self.coef_m_fine = coef_m_fine
self.phi1 = min_coverage
self.phi2 = 1 - max_coverage
def __call__(self, M):
loss_area = F.relu(self.phi1 - M.mean(dim=(-2,-1))).mean()
loss_inv_area = F.relu(self.phi2 - (1 - M).mean(dim=(-2,-1))).mean()
return self.coef_m_fine * (loss_area + loss_inv_area)